CN-122019133-A - Causal modeling method and device in computational power network, electronic equipment and storage medium

CN122019133ACN 122019133 ACN122019133 ACN 122019133ACN-122019133-A

Abstract

The invention provides a causal modeling method, a causal modeling device, electronic equipment and a storage medium in a computational power network, wherein the causal modeling method in the computational power network comprises the steps of collecting cross-domain multi-level computational power network data in the computational power network; extracting multi-dimensional characteristics in the power calculation network data, and screening based on the association degree of the multi-dimensional characteristics and faults to obtain multi-dimensional target characteristics; and constructing a cross-domain multi-level dynamic causal graph based on the computational power network topology structure and the multi-dimensional target characteristics. The invention provides a scheme for constructing a cross-domain multi-level dynamic causal graph of a unified adaptive computational power network, which can solve the core pain points of difficult cross-domain fault tracing, weak characteristic association, inaccurate fault positioning and the like in the computational power network based on the dynamic causal graph, and greatly improves the fault positioning accuracy in the complex computational power network.

Inventors

XU JUNDONG
SONG DEHUA
Zhai Yuehao

Assignees

浪潮通信信息系统有限公司

Dates

Publication Date: 20260512
Application Date: 20251224

Claims (10)

1. A causal modeling method in a computational power network, comprising: Acquiring cross-domain multi-level computing power network data in a computing power network; extracting multi-dimensional characteristics in the power calculation network data, and screening based on the association degree of the multi-dimensional characteristics and faults to obtain multi-dimensional target characteristics; And constructing a cross-domain multi-level dynamic causal graph based on the computational power network topology structure and the multi-dimensional target characteristics.
2. The causal modeling method in a computational power network according to claim 1, wherein said constructing a cross-domain multi-level dynamic causal graph based on computational power network topology and said multi-dimensional target features comprises: Constructing an initial causal graph architecture according to the computational power network topological structure, and constructing a causal subgraph based on the initial causal graph architecture layer; determining nodes in each causal subgraph based on the multi-dimensional target features; Determining a first causal relationship between nodes in the causal subgraph of the same layer based on the multi-dimensional target feature, and determining a second causal relationship between nodes in the causal subgraph of different layers; The dynamic causal graph is constructed based on the multi-dimensional target feature, the first causal relationship, and the second causal relationship.
3. The causal modeling method in the computational power network according to claim 1, wherein the extracting the multidimensional feature in the computational power network data and screening based on the association degree of the multidimensional feature and the fault to obtain the multidimensional target feature comprises: extracting multi-dimensional features in the computing power network data, wherein the multi-dimensional features comprise basic dimension features, topological association features, time sequence evolution features and cross-dimension coupling features; And calculating the maximum information coefficient of each multi-dimensional characteristic, and carrying out association screening based on the maximum information coefficient to obtain the multi-dimensional target characteristic.
4. The causal modeling method in a computational power network according to claim 1, wherein, after the constructing a cross-domain multi-level dynamic causal graph based on the computational power network topology and the multi-dimensional target features, the method further comprises: and if a causal change trigger event is detected, updating the dynamic causal graph based on the causal change trigger event.
5. The causal modeling method in a computational power network of claim 1, wherein the collecting computational power network data of a cross-domain, multi-level computational power network comprises: Acquiring the power calculation network data in the power calculation network according to the power calculation network topological structure; And performing time sequence calibration on the calculation network data.
6. The causal modeling method in a computational power network according to any of claims 1 to 5, wherein the method further comprises, after constructing a cross-domain multi-level dynamic causal graph based on computational power network topology and the multi-dimensional target features: and constructing a target hash index of each node in the dynamic causal graph, wherein the target hash index comprises a region code, a position code, a hierarchy code, a category code and an object identifier.
7. The causal modeling method in a computational power network according to any of claims 1 to 5, wherein the method further comprises, after constructing a cross-domain multi-level dynamic causal graph based on computational power network topology and the multi-dimensional target features: Acquiring fault information; and based on the fault information, performing root cause positioning in the dynamic causal graph to obtain a fault root cause result.
8. A causal modeling apparatus in a computational power network, comprising: The acquisition module is configured to acquire cross-domain multi-level computing power network data in the computing power network; the extraction module is configured to extract the multi-dimensional characteristics in the power calculation network data, and screen the power calculation network data based on the association degree of the multi-dimensional characteristics and faults to obtain multi-dimensional target characteristics; The first construction module is configured to construct a cross-domain multi-level dynamic causal graph based on the computational power network topology and the multi-dimensional target feature.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that the processor implements the causal modeling method in the computational power network according to any of claims 1 to 7 when executing the computer program.
10. A non-transitory computer readable storage medium, having stored thereon a computer program, which when executed by a processor implements a causal modeling method in a computational power network according to any of claims 1 to 7.

Description

Causal modeling method and device in computational power network, electronic equipment and storage medium Technical Field The present invention relates to the field of computing power networks, and in particular, to a causal modeling method, apparatus, electronic device, and storage medium in a computing power network. Background With the advancement of integrated power calculation networks, a plurality of layers such as a physical layer, a virtual layer, a scheduling layer, an application layer, a service layer and the like which cross regions exist in the power calculation networks, the object and the fault mode of each layer have huge differences, and meanwhile, the cross links are associated, so that great challenges are brought to the rapid positioning and processing of faults in the power calculation networks. The existing technology for fault location of the computational power network mainly relies on the traditional monitoring tool, a static rule base and fragmentation index analysis, focuses on the processing of single-domain problems, lacks cross-dimension cooperative capability and has serious fault location fragmentation. Meanwhile, with the rapid development of the technology of the intelligent agent, a new direction is brought to the automatic and intelligent operation and maintenance of the computational power network, however, hidden danger exists in the interpretability and the reliability of the intelligent agent, and a scheme which can supplement the intelligent agent and provide auxiliary fault positioning decisions for the intelligent agent is urgently needed. And at present, a unified computational power network causal modeling framework does not exist, so that the misjudgment rate of the decision of the operation and maintenance agent is high. Disclosure of Invention The invention provides a causal modeling method, a causal modeling device, electronic equipment and a storage medium in a computational power network, which are used for solving the defect that a uniform cross-level dynamic causal graph for adapting the computational power network does not exist in the prior art. The invention provides a causal modeling method in a computational power network, which comprises the following steps: Acquiring cross-domain multi-level computing power network data in a computing power network; extracting multi-dimensional characteristics in the power calculation network data, and screening based on the association degree of the multi-dimensional characteristics and faults to obtain multi-dimensional target characteristics; And constructing a cross-domain multi-level dynamic causal graph based on the computational power network topology structure and the multi-dimensional target characteristics. According to the causal modeling method in the computational power network provided by the invention, the construction of a cross-domain multi-level dynamic causal graph based on the computational power network topological structure and the multi-dimensional target characteristics comprises the following steps: Constructing an initial causal graph architecture according to the computational power network topological structure, and constructing a causal subgraph based on the initial causal graph architecture layer; determining nodes in each causal subgraph based on the multi-dimensional target features; Determining a first causal relationship between nodes in the causal subgraph of the same layer based on the multi-dimensional target feature, and determining a second causal relationship between nodes in the causal subgraph of different layers; The dynamic causal graph is constructed based on the multi-dimensional target feature, the first causal relationship, and the second causal relationship. According to the causal modeling method in the computational power network provided by the invention, the multi-dimensional characteristics in the computational power network data are extracted, and screening is performed based on the association degree of the multi-dimensional characteristics and faults to obtain multi-dimensional target characteristics, and the causal modeling method comprises the following steps: extracting multi-dimensional features in the computing power network data, wherein the multi-dimensional features comprise basic dimension features, topological association features, time sequence evolution features and cross-dimension coupling features; And calculating the maximum information coefficient of each multi-dimensional characteristic, and carrying out association screening based on the maximum information coefficient to obtain the multi-dimensional target characteristic. According to the causal modeling method in the computational power network provided by the invention, after the cross-domain multi-level dynamic causal graph is constructed based on the computational power network topological structure and the multi-dimensional target characteristics, the method further comprises the follo